Papers with model reasoning
Aligning LLMs for Multilingual Consistency in Enterprise Applications (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Large language models (LLMs) remain unreliable for global enterprise applications due to performance gaps between high-resource and mid/low-resourced languages . |
| Approach: | They propose a batch-wise alignment strategy that aligns model outputs across languages . this method improves non-English accuracy by up to 23.9% without compromising English performance . |
| Outcome: | The proposed approach improves non-English accuracy by up to 23.9% without compromising English performance, model reasoning, or retrieval quality. |
FactSearch: An Interactive Agentic Fact Search System for Verifying Large Language Model Outputs (2026.acl-demo)
Copied to clipboard
| Challenge: | Existing tool-augmented verification systems depend on opaque search APIs, introducing uncontrolled variability into factuality evaluation. |
| Approach: | They propose a reproducibility-oriented agentic fact search system for claim-level factuality verification built on a locally aggregated open-source search infrastructure. |
| Outcome: | The proposed system decomposes model outputs into atomic factual claims, generates targeted search queries, retrieves supporting evidence via a self-hosted meta-search engine, and performs modular verification within a fully configurable pipeline. |
TACO: Enhancing Multimodal In-context Learning via Task Mapping-Guided Sequence Configuration (2025.emnlp-main)
Copied to clipboard
| Challenge: | Multimodal in-context learning (ICL) is a key mechanism for harnessing the capabilities of large vision–language models. |
| Approach: | They propose a transformer-based model with task-aware attention that dynamically configures ICL sequences. |
| Outcome: | Experiments on five LVLMs and nine datasets show that TACO surpasses baselines across diverse ICL tasks. |
Analysing LLM Persona Generation and Fairness Interpretation in Polarised Geopolitical Contexts (2026.eacl-srw)
Copied to clipboard
| Challenge: | Large language models are increasingly used for social simulation and persona generation. |
| Approach: | They analysed personas generated for Palestinian and Israeli identities by popular LLMs across 640 experimental conditions, varying context and assigned roles. |
| Outcome: | The results show that large language models are increasingly utilised for social simulation and persona generation. |
Garbage In, Reasoning Out? Why Benchmark Scores are Unreliable and What to Do About It (2026.findings-eacl)
Copied to clipboard
| Challenge: | Using social reasoning benchmarks, we uncover pervasive flaws in both benchmark items and evaluation methodology. |
| Approach: | They audit three widely used social reasoning benchmarks and identify flaws in their design and evaluation methodology. |
| Outcome: | The results challenge the validity of current benchmark-based claims about social reasoning in large language models. |
uMedSum: A Unified Framework for Clinical Abstractive Summarization (2025.acl-long)
Copied to clipboard
Aishik Nagar, Yutong Liu, Andy T. Liu, Viktor Schlegel, Vijay Prakash Dwivedi, Arun-Kumar Kaliya-Perumal, Guna Pratheep Kalanchiam, Yili Tang, Robby T. Tan
| Challenge: | Clinical abstractive summarization struggles to balance faithfulness and informativeness, sacrificing key information or introducing confabulations. |
| Approach: | They develop a modular hybrid framework that integrates confabulation removal and key information addition into abstractive summarization methods. |
| Outcome: | The proposed framework outperforms state-of-the-art abstractive summarization methods in both quantitative metrics and expert evaluations. |
Beyond Examples: Towards Automated Thought-level In-Context Reasoning for Large Language Models (2026.acl-long)
Copied to clipboard
Jinyang Wu, Mingkuan Feng, Shuai Zhang, Feihu Che, Zhengqi Wen, Chonghua Liao, Ling Yang, Haoran Luo, Zheng Lian, Jianhua Tao
| Challenge: | In-context learning (ICL) struggles with complex reasoning due to superficial, example-level implicit imitation. |
| Approach: | They propose an automated method that shifts from surface-level examples to more guidance-oriented thought patterns. |
| Outcome: | The proposed method achieves 80.6% accuracy on MATH and 62.5% on AMC, surpassing GPT-4o’s 77.2% and 57.5% accuracy. |
Diagnosing Memorization in Chain-of-Thought Reasoning, One Token at a Time (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) perform well on reasoning benchmarks but often fail when inputs alter slightly, raising concerns about overreliance on memorization. |
| Approach: | They propose a framework for Source-aware Token-level Identification of Memorization which attributes each token in a reasoning chain to one of multiple memorization sources based on their statistical co-occurrence with the token in the pretraining corpus. |
| Outcome: | The proposed framework attributes each token in a reasoning chain to one of multiple memorization sources based on their statistical co-occurrence with the token in the pretraining corpus. |
TemplateRL: Structured Template-Guided Reinforcement Learning for LLM Reasoning (2026.findings-acl)
Copied to clipboard
Jinyang Wu, Chonghua Liao, Mingkuan Feng, Shuai Zhang, Zhengqi Wen, Haoran Luo, Ling Yang, Huazhe Xu, Jianhua Tao
| Challenge: | Existing RL methods rely on unstructured self-sampling to fit scalar rewards, resulting in inefficient rollouts. |
| Approach: | They propose a structured template-guided RL framework that augments policy optimization with explicit template guidance. |
| Outcome: | Experiments show that TemplateRL outperforms GRPO and GRPI by 99% on AIME and 41% on AMC with superior stability on weak models and remarkable cross-domain generalization. |
Self-contradictory reasoning evaluation and detection (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models (LLMs) have shown impressive reasoning ability, but many downstream reasoning tasks focus on performance-wise evaluation. |
| Approach: | They define and assess the Self-Contra rate across three datasets and delve into finer-grained categories of Self-contra reasoning. |
| Outcome: | The proposed model can detect self-contra reasoning with a 52.2% F1 score, much lower than for humans. |
How Chain-of-Thought Works? Tracing Information Flow from Decoding, Projection, and Activation (2026.findings-acl)
Copied to clipboard
| Challenge: | Chain-of-Thought (CoT) prompting significantly enhances model reasoning, yet its internal mechanisms remain poorly understood. |
| Approach: | They reversely traced information flow across decoding, projection, and activation phases and found that CoT may serve as a decoding space pruner . |
| Outcome: | The proposed framework can be used to design more efficient and robust prompts. |
Gender Bias in Decision-Making with Large Language Models: A Study of Relationship Conflicts (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models acquire beliefs about gender from training data and can therefore generate text with stereotypical gender attitudes. |
| Approach: | They use a decision-making lens to examine gender equity within large language models . they explore relationships through typical and gender-neutral names . |
| Outcome: | The proposed model generation and classification models exhibit stereotypical gender biases . the proposed model generates gender-neutral names, with and without safety enhancements, and egalitarian versus traditional scenarios across topics. |
FLamE: Few-shot Learning from Natural Language Explanations (2023.acl-long)
Copied to clipboard
| Challenge: | Recent work has shown limited utility of natural language explanations in improving classification. |
| Approach: | They propose a two-stage few-shot learning framework that generates explanations and fine-tunes a smaller model with generated explanations. |
| Outcome: | The proposed framework increases inference accuracy over strong baselines, but human evaluation reveals that the majority of generated explanations does not adequately justify classification decisions. |
T2DR: A Two-Tier Deficiency-Resistant Framework for Incomplete Multimodal Learning (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing incomplete multimodal learning frameworks are inadequate for integrating multimodal data. |
| Approach: | They propose a framework for incomplete multimodal learning that is deficiency-resistant and provides two modules to address fine-grained deficiencies. |
| Outcome: | The proposed framework outperforms the SOTA models on two well-known multimodal benchmarks. |
Self-Reflective Generation at Test Time (2026.acl-long)
Copied to clipboard
| Challenge: | Existing self-reflection mechanisms are reactive and inefficient for large language models . a fundamental tension persists between the ability to execute complex multi-step reasoning and the ability of the model to generate coherent traces. |
| Approach: | They propose a test-time framework that reflects before generating at uncertain points . SRGen utilizes dynamic entropy thresholding to identify high-uncertainty tokens . |
| Outcome: | The proposed framework can significantly strengthen large language models' reasoning process. |
Self-Polish: Enhance Reasoning in Large Language Models via Problem Refinement (2023.findings-emnlp)
Copied to clipboard
Zhiheng Xi, Senjie Jin, Yuhao Zhou, Rui Zheng, Songyang Gao, Jia Liu, Tao Gui, Qi Zhang, Xuanjing Huang
| Challenge: | Existing prompting methods have been used to enhance multistep reasoning capabilities of large language models, but they have overlooked the potential of formulating higher-quality problems. |
| Approach: | They propose a method that starts from the problem side and refines problems to be more comprehensible and solvable for models. |
| Outcome: | The proposed method achieves notable and consistent effectiveness on five reasoning benchmarks across different models. |
Beyond Benchmarks: A Capability-Based Maturity Model for Systematic AI Integration in Hospitals (2026.findings-acl)
Copied to clipboard
| Challenge: | Current Large Language Models (LLMs) excel in standardized tests focused on medical knowledge recall, but not in real-world healthcare scenarios. |
| Approach: | They propose a "capability-based hospital AI Maturity Model" framework that categorizes capabilities into distinct maturity levels . medical artificial intelligence is currently at a critical transition stage from technical verification to deep clinical integration . |
| Outcome: | The proposed model provides a clear, stepwise evolutionary path for hospitals from foundational infrastructure construction to ubiquitous intelligence. |
CaT-Bench: Benchmarking Language Model Understanding of Causal and Temporal Dependencies in Plans (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies on reasoning in plans focus on classical problems, simulated environments, or restricted language such as PDDL, but real-world plans cannot be tested to test for correctness and reliability. |
| Approach: | They propose a benchmark question that tests whether a step must necessarily occur before or after another in cooking recipe plans. |
| Outcome: | The proposed question-driven evaluation shows that SOTA LLMs are underwhelming and biased towards predicting dependence more often, but the best F1 result is 0.73. |
Decoupling Reasoning and Knowledge Injection for In-Context Knowledge Editing (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing knowledge editing approaches directly edit model context without isolating target knowledge from the reasoning path of model inference, resulting in unreliable and low-quality outputs, especially in multi-hop tasks. |
| Approach: | They propose a framework that separates model reasoning from knowledge editing and propose 'DecKER' that allows users to modify specific factual associations without retraining the entire model. |
| Outcome: | The proposed framework significantly improves multi-hop reasoning performance by mitigating knowledge conflicts and preserving reasoning integrity. |
Small Models Struggle to Learn from Strong Reasoners (2025.findings-acl)
Copied to clipboard
Yuetai Li, Xiang Yue, Zhangchen Xu, Fengqing Jiang, Luyao Niu, Bill Yuchen Lin, Bhaskar Ramasubramanian, Radha Poovendran
| Challenge: | a small learning gap exists between large and small language models . long CoT data and large model responses are not beneficial for small models - a problem that may be due to the small student model's ability to handle distribution shifts. |
| Approach: | They propose a mix distillation strategy that balances reasoning complexity by combining long and short CoT examples or reasoning from both larger and smaller models. |
| Outcome: | The proposed strategy outperforms training on large and small models on short CoT and small model CoT. |
Evo-PI: Aligning Medical Reasoning via Evolving Principle-Guided Supervision (2026.acl-long)
Copied to clipboard
| Challenge: | Existing models with static prompts, rules, or reward models are constrained by static supervision, which often fails to shape the underlying reasoning process, leading to brittle generalization and performance saturation in complex decision-making tasks. |
| Approach: | They propose a principle-centric learning framework that treats reasoning principles as explicit, language-based supervision signals that can be generated, evaluated, and iteratively evolved. |
| Outcome: | The proposed framework treats reasoning principles as explicit, language-based supervision signals that can be generated, evaluated, and iteratively evolved. |
Monotonic Scaffolding as a Diagnostic Lens for Legal Reasoning in LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Modern evaluation of Legal QA systems is shifting from terminal accuracy toward process-aware analyses of model reasoning. |
| Approach: | They propose a diagnostic framework grounded in monotonic scaffolding where language models receive gold-standard, case-relevant information across stages aligned with the canonical legal framework FIRAC. |
| Outcome: | The proposed framework evaluates 3,123 Brazilian Bar Exam questions . it shows that terminal accuracy overestimates legal reasoning competence . |